JP2020091232A - Medical image processor, nuclear medicine diagnosis device, method for processing medical image, and learning device - Google Patents

Abstract

To precisely estimate information on the coarctation of a coronary artery on the basis of a clear image of the inside state of a heart.SOLUTION: The medical image processor according to an embodiment includes a first acquisition unit and a coarctation information generation unit. The first acquisition unit acquires a nuclear medicine diagnosis image obtained by scanning the heart of a subject. The coarctation information generation unit generates information on the coarctation of a coronary artery of the subject by inputting at least the nuclear medicine diagnosis image acquired by the first acquisition unit into a learned model which outputs the information on the coarctation of the coronary artery of the subject on the basis of at least the nuclear medicine diagnosis image.SELECTED DRAWING: Figure 1

Description

Embodiments of the present invention relate to a medical image processing apparatus, a nuclear medicine diagnosis apparatus, a medical image processing method, and a learning apparatus.

A nuclear medicine diagnostic device for performing scintigraphy is known. The nuclear medicine diagnostic apparatus includes a SPECT (Single Photon Emission computed Tomography) apparatus and a PET (Positron Emission Tomography) apparatus. A nuclear medicine diagnostic device is distributed in a living body by utilizing the property that a drug such as a blood flow marker or a tracer containing a radioisotope (RI) is selectively taken into a specific tissue or organ in the living body. The gamma rays emitted from the RI are detected by a gamma camera having a gamma ray detector arranged outside the living body. The nuclear medicine diagnostic apparatus can provide an internal image of the subject (patient) by generating a nuclear medicine diagnostic image in which the dose distribution of gamma rays detected by the gamma ray detector is imaged.

The nuclear medicine diagnostic apparatus is mainly used for the head or heart of a subject. When inspecting the heart, a stress SPECT inspection is performed by a nuclear medicine diagnostic device, and a coronary CT inspection is performed when some abnormality is found. Since a coronary CT examination uses a contrast medium, it is necessary to consider the allergic constitution of the subject, and it is preferable to avoid it for patients with dialysis, diabetes, asthma, etc. as much as possible. For this reason, a load SPECT test is first performed on a subject having the above-mentioned attributes, and a coronary CT test is performed when it is determined that a mild perfusion abnormality or determination is difficult. However, in the related art, there are cases where it is not possible to accurately estimate the information regarding the stenosis of the coronary artery from the result of the examination by the nuclear medicine diagnostic apparatus such as the load SPECT examination, that is, the image showing the internal state of the heart.

U.S. Patent Application Publication No. 2017/0071479

The problem to be solved by the present invention is to accurately estimate information on stenosis of a coronary artery based on an image representing the internal state of the heart.

The medical image processing apparatus according to the embodiment has a first acquisition unit and a stenosis information generation unit. The first acquisition unit acquires a nuclear medicine diagnostic image obtained by scanning the heart of the subject. The stenosis information generation unit, by inputting the nuclear medicine diagnosis image acquired by at least the first acquisition unit to a learned model that outputs information regarding the stenosis of the coronary artery of the subject based on at least the nuclear medicine diagnosis image, Generating information about the stenosis of the coronary artery of the subject.

The block diagram which shows an example of the nuclear medicine diagnostic apparatus 1 which concerns on 1st Embodiment. FIG. 5 is a diagram schematically showing how a PolarMap is created. The figure which compared PolarMap at the time of load and at rest. The figure which shows an example of a structure of the learning device 100. The figure which shows the content of the process by the model generation function 154 typically. The figure which shows typically the content of the process by the stenosis information generation function 75. The figure which illustrated the image displayed on the display 54 as a processing result of the stenosis information generation function 75 (the 1). The figure which illustrated the image displayed on the display 54 as a processing result of the stenosis information generation function 75 (the 2). The figure which illustrated the image displayed on the display 54 as a processing result of the stenosis information generation function 75 (the 3). 9 is a flowchart showing an example of the flow of processing executed by the model generation function 154. The flowchart which shows an example of the flow of the process performed by the console device 50. The figure which shows an example of the structure of the medical service provision apparatus 200 which concerns on 2nd Embodiment, and the nuclear medicine diagnostic apparatus 1A.

Hereinafter, a medical image processing apparatus, a nuclear medicine diagnosis apparatus, a medical image processing method, and a learning apparatus according to embodiments will be described with reference to the drawings. In the following embodiments, the nuclear medicine diagnostic device is assumed to be a SPECT device, but the nuclear medicine diagnostic device is not limited to the SPECT device and may be a PET device.

(First embodiment)
FIG. 1 is a block diagram showing an example of a nuclear medicine diagnosis apparatus 1 according to the first embodiment. The nuclear medicine diagnostic device 1 includes, for example, a scanner device 10 and a console device 50. In the first embodiment, the console device 50 is an example of a medical image processing device.

[Scanner device]
The scanner device 10 includes, for example, a fixed mount 12, a rotary mount 14, a rotary drive device 16, three sets of gamma ray detectors 20 and collimators 22 mounted on the rotary mount 14 while being shifted by 120 degrees, and a collimator drive circuit. 24, a data collection circuit 26, a top plate 30, and a top plate drive device 32.

The fixed base 12 is fixed to the floor in the room where the scanner device 10 is installed. The rotary base 14 is supported rotatably around the rotation axis AX with respect to the fixed base 12. The subject P is placed on the top plate 30 so that the body axis thereof is substantially parallel to the rotation axis AX of the rotary mount 14.

The rotation drive device 16 rotates the rotary mount 14 around the rotation axis AX. The rotary drive device 16 has, for example, drive means such as a motor, electronic components for controlling the drive means, and transmission means such as rollers for transmitting the rotational force of the rotation shaft of the drive means to the rotary mount 14. The rotation driving device 16 is controlled by the processing circuit 70. For example, the processing circuit 70 collects the projection data of the subject P from a plurality of directions by rotating the gamma ray detector 20 around the rotation axis AX continuously or stepwise via the rotary mount 14. To enable.

The gamma ray detector 20 detects gamma rays emitted by RI (radioisotope) such as technesium administered to the subject P. The detection timing of the gamma ray detector 20 is controlled by the processing circuit 70. The gamma ray detector 20 is, for example, a scintillator type detector or a semiconductor type detector. This will be described later.

The collimator 22 regulates the incident angle of gamma rays that enter the gamma ray detector 20. The collimator 22 is formed of a substance such as lead or tungsten that is hard to transmit radiation. The collimator 22 is provided with a plurality of holes for controlling the direction in which photons fly. The cross section of the hole has, for example, a polygonal shape such as a hexagon.

When the gamma ray detector 20 is a scintillator type detector, for example, the gamma ray detector 20 emits a momentary flash when a gamma ray collimated by the collimator 22 is incident, a light guide, and a light emitted from the scintillator. It has a plurality of photomultiplier tubes arranged two-dimensionally for detecting the light and a scintillator electronic circuit. The scintillator is composed of, for example, thallium-activated sodium iodide NaI(Tl). The electronic circuit for the scintillator, based on the output of the multiple photomultiplier tubes, generates the information on the incident position of the gamma rays based on the output of the multiple photomultiplier tubes each time the event of the incident gamma rays occurs. (Position information), incident intensity information, and incident time information are generated and output to the processing circuit 70 of the console device 50. This position information may be information of two-dimensional coordinates in the detection surface, or the detection surface may be virtually divided into a plurality of divided areas (primary cells) in advance (for example, divided into 1024×1024 pieces). Alternatively, the information may be information indicating which primary cell has been incident.

When the gamma ray detector 20 is a semiconductor type detector, the gamma ray detector 20 has a plurality of two-dimensionally arranged semiconductor elements for detecting the gamma rays collimated by the collimator 22 and a semiconductor electronic circuit. . The semiconductor element is formed of, for example, CdTe or CdZnTe (CZT). The semiconductor electronic circuit generates incident position information, incident intensity information, and incident time information based on the output of the semiconductor element and outputs the information to the processing circuit 70 each time an event that a gamma ray is incident occurs. This position information is information indicating which semiconductor element among a plurality of semiconductor elements (for example, 1024×1024) has entered.

The collimator drive circuit 24 drives, for example, the gamma ray detector 20 and the collimator 22 in a direction in which the gamma ray detector 20 and the collimator 22 are moved closer to or farther from the rotation axis AX of the rotary mount 14.

The data acquisition circuit 26 includes, for example, a printed circuit board. The data acquisition circuit 26 executes imaging of the subject P by controlling at least the gamma ray detector 20 according to an instruction from the processing circuit 70. The data collection circuit 26 detects the detection position information of the gamma rays detected by the gamma ray detector 20, the intensity information, the information indicating the relative position between the gamma ray detector 20 and the subject P, and the detection time of the gamma ray as the incident event of the gamma ray. To be collected.

The top plate 30 mounts the subject P and regulates the movement of the subject P. The top plate driving device 32 is controlled by the processing circuit 70, and moves the top plate 30 along the rotation axis AX of the rotary mount 14 or in the vertical direction (Y direction in the drawing). The top driving device 32 has a motor as a drive source, electronic parts for controlling the motor, and the like.

[Console device]
The console device 50 includes, for example, an input interface 52, a display 54, a storage circuit 56, a network connection circuit 58, and a processing circuit 70. The console device 50 may be a device specifically designed for the nuclear medicine diagnostic device 1 or may be a general-purpose personal computer or workstation in which a necessary program is installed. In the former case, a part of the configuration of the console device 50 may be dispersedly arranged on the fixed base 12.

The input interface 52 includes devices such as a keyboard, a mouse, a touch panel, and a trackball. The input interface 52 outputs an operation input signal corresponding to a user operation to the processing circuit 70.

The display 54 is, for example, an LCD (Liquid Crystal Display) or an organic EL (Electroluminescence) display.

The storage circuit 56 includes a non-primary storage medium such as an HDD (Hard Disk Drive) and a flash memory. The storage circuit 56 may include a storage medium such as a RAM (Random Access Memory) or a register. The memory circuit 56 stores data that can be read by the hardware processor. The storage circuit 56 stores, for example, information such as a nuclear medicine diagnosis image 56-1 generated by the processing circuit 70, findings information 56-2, and a learned model 56-3. Further, the storage circuit 56 may store a program executed by the hardware processor of the processing circuit 70.

The network connection circuit 58 includes, for example, a network card having a printed circuit board, a wireless communication module, or the like. The network connection circuit 58 implements an information communication protocol according to the form of the network to be connected. The network includes, for example, a LAN (Local Area Network), a WAN (Wide Area Network), the Internet, a cellular network, a dedicated line, and the like.

The processing circuit 70 includes, for example, a scan control function 71, a preprocessing function 72, a reconstruction processing function 73, a PolarMap (polar map) creation function 74, and a stenosis information generation function 75. The processing circuit 70 realizes these functions by, for example, a hardware processor executing a program stored in the storage circuit 56.

The hardware processor is, for example, a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), an application specific integrated circuit (ASIC), a programmable logic device (for example, a simple programmable logic device (Simple Programmable Logic). Device; SPLD) or a complex programmable logic device (CPLD), field programmable gate array (Field Programmable Gate Array; FPGA)), and the like. Instead of storing the program in the memory circuit 56, the program may be directly incorporated in the circuit of the hardware processor. In this case, the hardware processor realizes the function by reading and executing the program incorporated in the circuit. The hardware processor is not limited to being configured as a single circuit, but may be configured as one hardware processor by combining a plurality of independent circuits to realize each function. Also, a plurality of constituent elements may be integrated into one hardware processor to realize each function.

The scan control function 71 controls a part or all of the rotation drive device 16, the collimator drive circuit 24, the data collection circuit 26, and the top plate drive device 32 based on the scan plan execution instruction received by the input interface 52. Then, the scan is executed to obtain the projection data from the data acquisition circuit 26.

The preprocessing function 72 executes preprocessing on the projection data based on the preprocessing condition information received by the input interface 52 or stored in the storage circuit 56, for example. The preprocessing includes, for example, uniformity correction processing, rotation center correction processing, preprocessing filter processing, processing for converting fan beam projection data into parallel beam projection data, and the like.

The reconstruction processing function 73 performs reconstruction processing on the projection data preprocessed by the preprocessing function 72 to generate volume data. The volume data is stored in the storage circuit 56 as one of the nuclear medicine diagnosis images 56-1. The nuclear medicine diagnosis image 56-1 is an image that reflects the amount of nuclide at each position of the myocardium. The reconstruction processing function 73 performs reconstruction processing based on, for example, the Chang method iterative approximation method (Iterative Chang). Instead of this, the reconstruction processing function 73 may perform the reconstruction processing based on the ML-EM (Maximum Likelihood-Expectation Maximization) method, the OS-EM (Ordered Subset Expectation Maximization) method, or the like.

The PolarMap creation function 74 creates a PolarMap (polar map) based on the volume data created by the reconstruction processing function 73. PolarMap is stored in the memory circuit 56 as another one of the nuclear medicine diagnosis images 56-1. PolarMap is generated by, for example, extracting a plurality of cross-sectional images such as a short-axis cross section, a vertical long-axis cross section, and a horizontal long-axis cross section from the volume data along the axis, and superimposing the extracted cross-sectional images. FIG. 2 is a diagram schematically showing how a PolarMap is created. The cross section CS in the drawing is a short-axis cross section. In the first embodiment, the PolarMap creation function 74 is an example of the first acquisition unit. The PolarMap creation function 74 may generate a PolarMap by extracting a vertical major axis cross section or a horizontal major axis cross section and superposing them instead of the minor axis cross section.

The process of acquiring projection data, generating volume data, and creating PolarMap is executed, for example, when the subject P is at rest and under load (load SPECT examination). The time of load refers to a state in which the heart is under load. In this state, the subject P should be exercised by a device such as a treadmill or an exercise bike (registered trademark), or a drug should be administered to the subject P so that the heart is artificially loaded. Produced in. FIG. 3 is a diagram comparing PolarMap under load and at rest. As shown in the figure, it can be seen that a portion with low blood flow (ischemic portion) is shown in PolarMap during loading. The distribution of the ischemic portion and the degree of decrease in blood flow are clues for estimating the stenosis of the coronary artery.

The stenosis information generation function 75 inputs information about the stenosis of the coronary artery of the subject P by inputting the PolarMap at rest, the PolarMap at the time of loading, and the doctor's findings information for these PolarMaps to the learned model 56-3. To generate. Details of the finding information will be described later. The learned model 56-3 is a model learned by machine learning so that the information about the stenosis of the coronary artery of the subject P is output by inputting these pieces of information.

[Learning device]
The learned model 56-3 is generated by, for example, a learning device separate from the nuclear medicine diagnostic device 1. Here, the learning device will be described. FIG. 4 is a diagram illustrating an example of the configuration of the learning device 100. The learning device 100 includes, for example, a network connection circuit 110, an input interface 120, a display 130, a storage circuit 140, and a processing circuit 150.

The network connection circuit 110 includes, for example, a network card having a printed circuit board, a wireless communication module, or the like. The network connection circuit 110 implements an information communication protocol according to the form of the network to be connected. The network includes, for example, LAN, WAN, Internet, cellular network, leased line and the like. By connecting the network connection circuit 58 of the nuclear medicine diagnosis apparatus 1 and the network connection circuit 110 of the learning apparatus 100 to the same network, the nuclear medicine diagnosis apparatus 1 and the learning apparatus 100 can communicate with each other.

The input interface 120 includes devices such as a keyboard, a mouse, a touch panel, and a trackball, for example. The input interface 120 outputs an operation input signal corresponding to a user operation to the processing circuit 150.

The display 130 is, for example, an LCD (Liquid Crystal Display) or an organic EL (Electroluminescence) display.

The storage circuit 140 includes a non-primary storage medium such as an HDD or a flash memory. The storage circuit 140 may also include a storage medium such as a RAM or a register. The memory circuit 140 stores data that can be read by the hardware processor. The storage circuit 140 stores information such as the learning data 140-1 and the teacher data 140-2 referred to by the processing circuit 150, and the learned model 140-3 generated by the processing circuit 150. Further, the storage circuit 140 may store a program executed by the hardware processor of the processing circuit 150.

The processing circuit 150 includes, for example, a data acquisition function 152 and a model generation function 154. The processing circuit 150 realizes these functions, for example, by a hardware processor executing a program stored in the storage circuit 140.

The data acquisition function 152 acquires, for example, PolarMaps at rest and under load and the doctor's findings information on these PolarMaps, and stores them in the storage circuit 140 as learning data 140-1. Further, the data acquisition function 152 acquires information on coronary stenosis related to the subject P related to the above learning data, and stores it in the storage circuit 140 as teacher data 140-2. A set of learning data 140-1 and teacher data 140-2 obtained from the same subject P are associated with each other. The data acquisition function 152 acquires these pieces of information from, for example, one or more nuclear medicine diagnosis apparatuses 1 via a network. Further, the data acquisition function 152 may acquire the data stored in the portable storage medium via a drive device (not shown). The data acquisition function 152 is an example of a second acquisition unit.

The finding information is, for example, a diagnosis result (estimation result) by a doctor who visually recognizes PolarMap at rest and under load. The doctor who visually recognizes PolarMap at rest and during load describes information such as “high stenosis of the right coronary artery” in the diagnostic imaging report or the medical chart, and inputs it using the input interface 52. Further, the finding information may be a set of evaluation values obtained by a doctor evaluating PolarMap at rest and under load for each of a plurality of divided areas. Further, the finding information is not limited to information created by a doctor, and may be information generated by image processing on PolarMap at rest and under load. In this case, for example, the stenosis information generation function 75 performs image processing on the PolarMap at rest and during load to generate finding information.

The information on the stenosis of the coronary artery includes, for example, information on the stenosis rate and information on the stenotic blood vessel site. The information on the stenosis rate and the information on the stenotic blood vessel region are results obtained by being judged by a doctor or the like, or obtained by computer processing based on the coronary CT, for example. These pieces of information may be transferred from information written in a medical record as a paper medium or electronic information by a doctor, or may be electronically acquired a diagnosis result of coronary CT. Further, the information regarding the stenosis of the coronary artery may be information indicating the presence or absence of blood vessels having a stenosis rate of 30% or more, or a CT-FFR (Fractional Flow Reserve) value. Good.

FIG. 5 is a diagram schematically showing the content of processing by the model generation function 154. The model generation function 154 inputs a plurality of sets of learning data 140 to a machine learning model in which connection information and the like are defined in advance and parameters such as connection coefficients are provisionally set, and the result is the learning data. The parameters in the machine learning model are adjusted so as to approach the teacher data 140-2 corresponding to 140. The model generation function 154 adjusts the parameters of the machine learning model by back propagation (back error propagation method), for example. The machine learning model is, for example, a DNN (Deep Neural Network) using a CNN (Convolution Neural Network). One set of learning data 140 is, for example, a pixel value of each pixel of PolarMap at rest, a pixel value of each pixel of PolarMap at load, and information that digitizes finding information (concatenet), It is in vector or matrix form. The information about the stenosis of the coronary artery is also represented by, for example, a numerical value of a predetermined dimension. The model generation function 154 terminates the process when backpropagation is performed on the training data 140-1 corresponding to the predetermined number of sets of learning data 140-1. The machine learning model at that time is the learned model 140-3. The model generation function 154 may generate the learned model 140-3 using only PolarMap, and does not input both the PolarMap at rest and the PolarMap at load, but only the PolarMap at rest. (Alternatively, the learned model 140-3 may be generated using the PolarMap at rest and the finding information).

The output of the learned model 140-3 is not limited to information that completely matches the teacher data 140-2, for example, information indicating the presence or absence of an event (that is, zero or one), and may be represented by a probability distribution. For example, the output of the learned model 140-3 may be information such as “the probability of stenosis occurring in the right coronary artery is 85%”. By arbitrarily setting the output layer of the learned model 140-3, it is possible to output information that has the same meaning as the teacher data 140-2 but does not completely match.

The data structure or program that functions as the learned model 140-3 may be stored as the learned model 56-3 in the storage circuit 56 of the nuclear medicine diagnostic device 1 at the time of sale of the nuclear medicine diagnostic device 1. It may be installed as a learned model 56-3 in the storage circuit 56 of the nuclear medicine diagnostic device 1 after sale. In the latter case, the learned model 140-3 is transmitted from the network connection circuit 110 of the learning device 100 to the network connection circuit 58 via the network and stored in the storage circuit 56, for example. The learned model 140-3 is stored in a portable storage device, and the learned model is stored in the storage circuit 56 by mounting the portable storage device on a drive device (not shown) of the nuclear medicine diagnostic apparatus 1. It may be installed as 56-3.

[Re: console device]
Hereinafter, the console device 50 will be described again. The stenosis information generation function 75 inputs the input data to the learned model 56-3 generated by the learning device 100 as described above, thereby generating information on the stenosis of the coronary artery of the subject P. FIG. 6 is a diagram schematically showing the content of processing by the stenosis information generation function 75. The stenosis information generation function 75 inputs, into the learned model 56-3, a combination of the PolarMap at rest, the PolarMap at load, and the finding information regarding the subject P as a target, as input data. Then, the result is output based on the information about the stenosis of the coronary artery output by the learned model 56-3. The stenosis information generation function 75 causes the display 54 to display an image showing the result, for example. Note that the stenosis information generation function 75 may input only PolarMap to the trained model 56-3, or may not input both PolarMap at rest and PolarMap at load, but only PolarMap at rest ( Alternatively, the resting PolarMap and finding information) may be input to the learned model 56-3.

7 to 9 are diagrams exemplifying an image displayed on the display 54 as a processing result of the stenosis information generating function 75. As shown in the figure, “probability of stenosis of each coronary artery” may be displayed as the processing result of the stenosis information generating function 75, or “stenosis rate prediction for coronary arteries with possible stenosis”. “Value” may be displayed, “Probability that the stenosis rate for each coronary artery is 30[%] or more” may be displayed, and “Probability that a coronary artery has a stenosis rate of 30[%] or more” is displayed. It may be displayed.

[Processing flow]
FIG. 10 is a flowchart showing an example of the flow of processing executed by the model generation function 154. First, the model generation function 154 acquires a set of learning data 140-1 (step S200).

Next, the model generation function 154 inputs the acquired one set of learning data 140-1 into the machine learning model (step S202) and subtracts an error from the teacher data 140-2 corresponding to the one set of learning data 140-1. Back propagation is performed (step S204).

Next, the model generation function 154 determines whether or not the processes of steps S202 and S204 have been performed on the predetermined set number of learning data 140-1 (step S206). When the processes of steps S202 and S204 have not been performed for the predetermined number of sets of learning data 140-1, the model generation function 154 returns the process to step S200. When the processes of steps S202 and S204 are performed on the predetermined set number of learning data 140-1, the model generation function 154 determines the learned model using the parameters at that time (step S208), and the process of this flowchart is performed. finish.

FIG. 11 is a flowchart showing an example of the flow of processing executed by the console device 50. First, the scan control function 71 controls each part of the nuclear medicine diagnosis apparatus 1 so as to perform the load SPECT examination (step S300).

Next, the PolarMap creation function 74 creates PolarMap at rest and under load (step S302). Next, the stenosis information generation function 75 acquires finding information based on PolarMap, for example, via the input interface 52 (step S304), generates input data, and inputs it to the learned model 56-3 (step S306). Then, the stenosis information generation function 75 outputs the result based on the output of the learned model 56-3 (step S308).

According to the first embodiment described above, PolarMap and the finding information are input to the learned model to generate information about the stenosis of the coronary artery of the subject P, and based on the image representing the internal state of the heart. Information about coronary stenosis can be accurately estimated.

(Second embodiment)
The second embodiment will be described below. In the first embodiment, the console device that is a part of the nuclear medicine diagnostic device is an example of the medical image processing device, but in the second embodiment, the medical image processing device is not a nuclear medicine diagnostic device. It is assumed that the medical image processing device and the learning device are separately configured and integrated. Hereinafter, this is referred to as a medical service providing device.

FIG. 12 is a diagram showing an example of the relationship between the configuration of the medical service providing apparatus 200 according to the second embodiment and the nuclear medicine diagnostic apparatus 1A. The nuclear medicine diagnostic device 1A includes a scanner device 10 and a console device 50A. The console device 50A has a configuration in which the stenosis information generating function 75 is removed from the console device 50 of the first embodiment. Further, the learned model 56-3 may not be stored in the storage circuit 56 of the console device 50A.

The medical service providing apparatus 200 includes, for example, a network connection circuit 210, an input interface 220, a display 230, a storage circuit 240, and a processing circuit 250.

The network connection circuit 210 has the same configuration, function, and role as the network connection circuit 110 in the first embodiment. The input interface 220 has the same configuration, function, and role as the input interface 120 in the first embodiment. The display 230 has the same configuration, function, and role as the display 130 in the first embodiment.

The storage circuit 240 includes a non-primary storage medium such as an HDD or a flash memory. The storage circuit 240 may also include a storage medium such as a RAM or a register. The memory circuit 240 stores data that can be read by the hardware processor. The storage circuit 240 stores information such as the learning data 240-1 and the teacher data 240-2 referred to by the processing circuit 250, and the learned model 240-3 generated by the processing circuit 250. Further, the storage circuit 240 may store a program executed by the hardware processor of the processing circuit 250.

The processing circuit 250 includes, for example, a data acquisition function 252, a model generation function 254, and a stenosis information generation function 256. The processing circuit 250 realizes these functions, for example, by a hardware processor executing a program stored in the storage circuit 240.

The data acquisition function 252 has the same function and role as the data acquisition function 152 in the first embodiment. Furthermore, the data acquisition function 252 in the second embodiment acquires input data such as PolarMap and finding information of the subject P to be examined from the nuclear medicine diagnostic apparatus 1A. In the second embodiment, the data acquisition function 252 is an example of the first acquisition unit and also an example of the second acquisition unit.

The model generation function 254 has the same function as the model generation function 254 in the first embodiment.

The stenosis information generating function 256 has the same function as the stenosis information generating function 75 in the first embodiment arranged on the medical service providing apparatus 200 side. The stenosis information generation function 256 inputs the input data acquired by the data acquisition function 252 to the learned model 240-3 generated by the model generation function 254, thereby generating information regarding the stenosis of the coronary artery of the subject P. To do. Then, the stenosis information generation function 256 transmits the generated information on the stenosis of the coronary artery of the subject P to the nuclear medicine diagnostic apparatus 1A by controlling the network connection circuit 210. On the side of the nuclear medicine diagnosis apparatus 1A, the display 54 displays the images illustrated in FIGS. 7 to 9 based on the received information.

According to the second embodiment described above, the same effect as that of the first embodiment can be obtained. Further, according to the second embodiment, by consolidating the stenosis information generation function in a limited place, it is possible to efficiently arrange the computer resources.

As described above, the arrangement of the nuclear medicine diagnostic device, the stenosis information generating function, and the learning function can be arbitrarily determined. For example, the device for inputting the input data to the learning model and outputting the information regarding the stenosis of the coronary artery and the learning device may exist separately from the nuclear medicine diagnostic device.

The input data input to the learning model by the medical image processing apparatus is not limited to nuclear medicine diagnostic images such as PolarMap, but may be any image as long as it is an image representing the internal state of the heart of the subject.

The embodiment described above can be expressed as follows.
A hardware processor,
A storage device storing a program,
When the hardware processor executes the program stored in the storage device,
Obtain a nuclear medicine diagnostic image obtained by scanning the heart of the subject,
By inputting at least the acquired nuclear medicine diagnosis image to a learned model that outputs information about the stenosis of the coronary artery of the subject based on at least the nuclear medicine diagnosis image, information on the stenosis of the coronary artery of the subject is generated. To do
The medical image processing apparatus configured as described above.

According to at least one embodiment described above, a first acquisition unit (74, 252) that acquires a nuclear medicine diagnostic image obtained by scanning the heart of the subject P, and a subject based on at least the nuclear medicine diagnostic image. Stenosis information generation for generating information on coronary stenosis of the subject P by inputting at least a nuclear medicine diagnostic image to a learned model (56-3, 240-3) that outputs information on coronary stenosis of the subject By having the section (75, 256), it is possible to accurately estimate the information on the stenosis of the coronary artery based on the image showing the internal state of the heart.

Although some embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the spirit of the invention. These embodiments and their modifications are included in the invention described in the claims and the equivalents thereof as well as included in the scope and the gist of the invention.

1, 1A Nuclear medicine diagnostic device 10 Scanner device 50, 50A Console device 52, 120, 220 Input interface 54, 130, 230 Display 56, 140, 240 Storage circuit 56-1 Nuclear medicine diagnostic image 56-2 Finding information 56-3 , 140-3 Learned models 58, 110, 210 Network connection circuits 70, 150, 250 Processing circuit 71 Scan control function 72 Preprocessing function 73 Reconstruction processing function 74 PolarMap creation function 75, 256 Stenosis information generation function 100 Learning device 140 -1 Learning data 140-2 Teacher data 152, 252 Data acquisition function 154, 254 Model generation function 200 Medical service providing device

Claims (9)

A first acquisition unit for acquiring a nuclear medicine diagnostic image obtained by scanning the heart of the subject;
By inputting at least the nuclear medicine diagnostic image acquired by the first acquisition unit to a learned model that outputs information regarding the stenosis of the coronary artery of the subject based on at least the nuclear medicine diagnostic image, A stenosis information generation unit that generates information about stenosis;
A medical image processing apparatus comprising: The nuclear medicine diagnostic image includes a nuclear medicine diagnostic image at rest of the subject, and a nuclear medicine diagnostic image when the subject is under load,
The stenosis information generation unit, by inputting at least the nuclear medicine diagnostic image at rest of the subject, and the nuclear medicine diagnostic image at the time of load of the subject to the learned model, the coronary artery of the subject. Generate information about stenosis,
The medical image processing apparatus according to claim 1. The learned model is a model for outputting information about the stenosis of the coronary artery of the subject based on the nuclear medicine diagnostic image and the finding information of a doctor or a radiologist for the nuclear medical diagnostic image,
The stenosis information generation unit inputs the nuclear medicine diagnostic image and the finding information to the learned model to generate information on stenosis of the coronary artery of the subject,
The medical image processing apparatus according to claim 1. The learned model is a model for outputting information regarding the stenosis of the coronary artery of the subject based on the nuclear medicine diagnosis image and the evaluation result of the nuclear medicine diagnosis image evaluated by a predetermined reference,
The stenosis information generation unit generates information on stenosis of the coronary artery of the subject by inputting the nuclear medicine diagnostic image and the evaluation result to the learned model.
The medical image processing apparatus according to any one of claims 1 to 3. The nuclear medicine diagnostic image is an image that reflects the amount of nuclide at each position of the myocardium,
A medical image processing apparatus according to any one of claims 1 to 4, A medical image processing apparatus according to any one of claims 1 to 5,
A scanner device for scanning the heart of the subject,
A nuclear medicine diagnostic device comprising: A medical image processing method executed by a computer, comprising:
Acquiring a nuclear medicine diagnostic image obtained by scanning the heart of the subject,
By inputting at least the acquired nuclear medicine diagnostic image to a learned model that outputs information on the coronary artery stenosis of the subject based on at least the nuclear medicine diagnostic image, information on the stenosis of the coronary artery of the subject is generated. That
A medical image processing method comprising: A second acquisition unit for acquiring a nuclear medicine diagnostic image obtained by scanning the heart of the subject and information regarding the stenosis of the coronary artery of the subject;
At least the nuclear medicine diagnostic image is used as learning data, and the information regarding the stenosis of the coronary artery of the subject is used as teacher data to generate a learned model that outputs information regarding the stenosis of the coronary artery of the subject based on at least the nuclear medicine diagnostic image. A model generator,
A learning device equipped with. A first acquisition unit that acquires at least an image representing the internal state of the heart of the subject;
A stenosis information generation unit that generates information about the stenosis of the coronary artery of the subject by inputting at least the image to a learned model that outputs information about the stenosis of the coronary artery of the subject based on at least the image,
A medical image processing apparatus comprising:

Images